BACKGROUND OF THE INVENTION
[0001] The present invention relates to a real-time, high-resolution cerebral biopotential
analysis system and method, and more particularly to a computer-based biopotential
diagnostic system and method for quantitatively determining, in a noninvasive manner,
cerebral phenomena that can be ascertained by analyzing the properties of cerebral
electrical activity.
[0002] Despite a considerable expenditure of time and effort, current approaches to the
quantitative, noninvasive assessment of cerebral electrical activity, as displayed
in an electroencephalographic (EEG) waveform, have not been successful in fully extracting
all of the information which is present in this complex waveform. A great need remains
for an accurate, sensitive, reliable, and practical neurological profiling technology.
In particular, contemporary intraoperative EEG monitoring techniques have not been
widely adopted due to their inherent limitations. Indeed, large numbers of medical
malpractice suits are believed to be related to post-anesthesia morbidity and mortality,
and if such EEG monitoring techniques were reliable they certainly would have been
adopted.
[0003] A number of devices known in the prior art are capable of tracking cerebral activity
qualitatively. Techniques involving the use of the classical conventional analog EEG
are restricted to analyses in the time domain, and require considerable training for
adequate interpretation. Moreover, since the resolution of the human eye at standard
EEG tracing speeds is limited, much of the fine structure of the EEG is invisible.
Thus, visual EEG assessment is better characterized as an art rather than a science.
[0004] The use of frequency (power spectrum) analysis of the EEG in the 1960s introduced
the notion of some basic processing of the signal prior to visual inspection and led
to the application of frequency analysis of the EEC to various cerebral monitoring
problems. In the past 25 years, over 100 papers have been published in the medical
literature describing applications of power spectral analysis for purposes such as
assessing the depth of anesthesia and cerebral ischemia under various intraoperative
conditions. United States patent No. 4,557,270 issued to John also describes the use
of power spectrum analysis to evaluate cerebral perfusion during open heart surgery.
Several recent studies, however, have shown many deficiencies in the use of power
spectral analysis to monitor cerebral perfusion and to determine postoperative neurological
outcome. In addition, neither power spectrum analysis nor any other monitoring technique
has been shown to be reliable, demonstrated by the fact that the Harvard Medical School
Anesthesia Monitoring Standard does not include any type of intraoperative neurological
monitoring, due, in all likelihood, to the complexity of interpreting raw EEC data
and the unreliability of existing automated systems utilizing power spectrum or time-domain
analytic techniques.
[0005] The discharge of thousands of bioelectrically active cells in the brain, organized
in larger, interacting neural centers contributes to the formation of an electrical
signal with a wide frequency spectrum that is rich in harmonics and extremely complex
dynamics. Embedded in that signal is information regarding frequency content, nonlinearities,
and phase relationships arising from the complex neuronal firing patterns that take
place. Such firing patterns change constantly making the statistical properties of
the EEC signal highly nonstationary. Because of the complexity of the EEC signal,
conventional time and frequency modes of analysis have not been able to fully profile
its behavior. This may be one of the reasons for the limited success of such approaches.
[0006] In the Fourier transform of the second order autocorrelation function (the power
spectrum), processes are represented as a linear summation of statistically-uncorrelated
sine-shaped wave components. Contemporary approaches to monitoring the EEC by means
of the power spectrum have thus suppressed information regarding nonlinearities and
inter-frequency phase relationships and are of limited utility in representing the
EEC's dynamic structure.
[0007] Because the EEG is highly dynamic and nonlinear, the phase relationships within the
EEG are the elements most likely to carry diagnostic information regarding cerebral
function. The Fourier transform of the third order autocorrelation function, or autobispectrum,
is an analytic process that quantifies deviation from normality, quadratic nonlinearities
and inter-frequency phase relationships within a signal. The Fourier transform of
the third order cross-correlation function, or cross bispectrum, is an analytic process
that provides similar information for two signals. We can generalize these techniques
by defining the Fourier transform of the nth-order auto/cross correlation function,
or the n-1 order auto/cross spectrum, as an analytic process that contains information
regarding deviation from normality, as well as n-1 order nonlinearities and inter-frequency
phase relationships in a signal. Auto/cross spectra beyond the bispectrum will be
referred to as higher-order spectra.
[0008] Autobispectrum analysis techniques have been applied to the EEC signal to demonstrate
the basic bispectral properties of the conventional EEC. Such studies have also been
conducted to search for differences between the waking and sleeping states. Autobispectrum
analysis and power spectrum analysis have also been used in an attempt to show that
the EEGs of monozygotic twins are similar in structure. United States Patents No.
4,907,597 and 5,010,891 issued to Chamoun describe the use of auto/cross bispectrum
analysis of the EEG to evaluate cerebral phenomena such as quantifying depth and adequacy
of anesthesia, pain responses induced by surgical stress, cerebral ischemia, consciousness,
degrees of intoxication, ongoing cognitive processes and inter-hemispheric dynamic
phase relations.
[0009] To date, no one has used auto higher-order spectrum or cross higher-order spectrum
analysis for neurological diagnoses or monitoring of the cerebral phenomena described
above.
[0010] A common problem in analyzing the data generated by any of the spectral techniques
discussed above is the fact that the EEG's frequency distribution may dramatically
change under relatively stable physiological conditions. Such changes will lead to
changes in the power spectrum, bispectrum, and higher order spectra at the corresponding
frequencies. For example, when hypnotic anesthetic agents are administered in low
to medium concentrations, there is a substantial increase in the EEG activity in the
12-18 Hz frequency band. High doses of the same agents will lead to a sudden reduction
in activity in the 12-18 Hz band and increase in activity in the 0.5-3.5 Hz band,
followed by burst suppression at extremely high concentrations. A frequency-based
analysis that uses the 12-18 Hz frequency band to track the patient's anesthetic depth
during the administration of a hypnotic agent will provide a misleading assessment
of the patient's depth when the shift in activity from high to low frequency occurs.
Such transitions are even more complicated when a mixture of anesthetic agents is
used.
[0011] Therefore, a principal object of the present invention is to provide a noninvasive,
high resolution electroencephalographic system and method capable of recognizing and
monitoring physical phenomena that are reflected in properties of cerebral electrical
activity.
[0012] Another object of the present invention is to provide a noninvasive electroencephalographic
system and method capable of determining and monitoring depth and adequacy of anesthesia,
pain responses during surgical stress, cerebral ischemia, cerebral hypoxia, levels
of consciousness, degrees of intoxication, altered evoked potential responses, and
normal or abnormal cognitive processes including but not limited to identifying patients
with Alzheimer's disease and HIV-related dementias.
[0013] The values of auto/cross power spectrum, auto/cross bispectrum, and auto/cross higher-order
spectrum arrays change with different interventions or disease states. Therefore,
these values are used to create a diagnostic criterion. The power spectrum, bispectrum,
and higher-order spectrum arrays are used to create a clinically useful single-valued
index. This index is expected to accurately portray the particular diagnostic determination
in question. The system uses these indices as a diagnostic figure of merit for the
assessment of depth and adequacy of anesthesia, pain responses during surgical stress,
cerebral ischemia, cerebral hypoxia, levels of consciousness, degree of intoxication,
altered evoked potential responses, and normal or abnormal cognitive processes including
but not limited to Alzheimer's disease and HIV-related dementias. This approach makes
it possible for any operator to meaningfully interpret the output of the diagnostic
device.
[0014] In situations where continuous monitoring is required, indices can be continuously
displayed on a video terminal, thereby enabling the operator to interactively evaluate
regions of interest. For record-keeping purposes, index values and other pertinent
variables can be sent to a hard copy output device or stored on a storage device.
[0015] These and other objects and features of the present invention are more fully explained
by the following detailed description and figures.
BRIEF DESCRIPTION OF THE FIGURES
[0016]
Fig. 1 is a schematic view of the system of the present invention for detecting cerebral
phenomena in a non-invasive manner;
Fig. 2 is a schematic view of a 19 channel EEG data acquisition and analysis system
utilized in the system of Fig. 1;
Fig. 3 is a schematic view of the microcomputer used to display the EEG power spectrum
and bispectrum higher-order spectrum in the system of Fig. 1;
Fig. 4 is a schematic view of the processing operations performed by the system of
Fig. 1;
Fig. 5 is a flow chart of the operations of the monitor module shown in Fig. 4;
Figs. 6(a) - 6(c) are views of sample display representations of diagnostic index
generated by the system of Fig. 1;
Fig. 7 is a flow chart of the operations of the acquisition and EEG raw data management
module of the system shown in Fig. 4;
Fig. 8 is a flow chart of the frequency-domain-based method for producing autobispectrum,
cross bispectrum, auto power spectrum, or cross power spectrum used by the system
of Fig. 1;
Fig. 9 is a flow chart of the parametric based method for producing autobispectrum,
cross bispectrum, auto power spectrum, or cross power spectrum in the system of Fig.
1;
Fig. 10(a) is a graph showing a bispectral density array generated by the system of
Fig. 1;
Fig. 10(b) is a graph showing a biphase array generated by the system of Fig. 1;
Fig. 10(c) is a graph showing a bicoherence array generated by the system of Fig.
1;
Fig. 10(d) is a graph showing an array of square root of real triple product generated
by the system of Fig. 1;
Fig. 11 is a flow chart of the operations of the diagnostic index derivation module
shown in Fig. 4;
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0017] Referring to Fig. 1, the apparatus of the present invention includes a 19 channel
EEG data acquisition and analysis system 12 connected to a microcomputer 18.
[0018] The EEC leads are connected to a patient's head 14 by a set of surface electrodes
13. The international 10/20 electrode system is preferred. The EEG signals are detected
by the electrodes and transmitted over a patient cable 16 to the EEG data acquisition
and analysis system 12.
[0019] The data acquisition and analysis system 12 filters and amplifies the EEG waveforms.
Commonly used digital signal processing (DSP) techniques are applied to digitize,
to low-pass filter (50 Hz), and to decimate the signals. Power spectral, bispectral,
and higher-order spectral processing can then be performed.
[0020] The system 12 generates all power spectrum, bispectrum, and higher-order spectrum
arrays. These arrays are then used in conjunction with clinically predetermined coefficient
arrays to produce diagnostic indices. These indices are sent to the host computer
18 and are displayed on the graphics display 20. Printed output of the diagnostic
index is also available on the hard copy output device 22 which is connected to the
microcomputer 18. The operator interacts with the acquisition and analysis components
of the system by means of a user input device 24 with feedback on the graphics display
20.
[0021] The 19 channel data acquisition and analysis system 12 is shown in greater detail
in Fig. 2. The EEC surface potential, detected by surface electrodes 13 mounted on
the patient's head 14, passes through an electrosurgery protection circuit 30, a defibrillator
protection circuit 32, and an amplifier/filter circuit 36 before being passed on to
the multi-channel analog to digital converter 38.
[0022] The electrosurgery protection circuit 30 includes a radio frequency (rf) filter,
which limits the rf current through the patient leads 16 to less than 100 microamperes
and thus protects the patient 15 from rf burns and protects the amplifiers 36 from
damage resulting from exceeding the absolute maximum input voltage specified by the
manufacturer. This circuit can be an LC circuit consisting of a generic inductor connected
in series to a generic capacitor which is then connected to ground.
[0023] The defibrillator protection circuit 32 limits the voltage to the amplifiers 36 to
a safe level when a defibrillator is applied to the patient 15 and discharged. This
circuit can consist of a generic resistor connected, in series with the signal path,
to a neon light bulb or other surge suppression device which is then connected to
ground.
[0024] The amplifier/filter circuitry 36 is controlled by the microcomputer 18 for gain
and filtering levels which may be adjusted by the operator. Preferred gain and filtering
settings are discussed below. This circuit section consists of three stages. The first
is a pre-amplifier stage that can be assembled using a wide variety of high-impedance
pre-amplifiers such as those sold by National Semiconductor, Sunnyvale CA. The second
is a stage composed of programmable filters which will allow an adjustable band pass
cutoff to be selected anywhere in the range of 0.1 Hz to 4 KHz. The filters can be
designed using components from Frequency Devices, Haverhill MA. The third stage is
composed of programmable amplifiers which can be assembled from operational amplifiers
used in conjunction with a multiplying digital to analog (D/A) converter. Both components
can be obtained from National Semiconductor. The multiplying D/A is used to set the
gain to the appropriate levels requested by the microcomputer 18.
[0025] The high impedance pre-amplifier of each channel will saturate to either the positive
or negative supply voltage if the input of the pre-amplifier is not terminated. This
will lead to large positive value or a large negative value at the output of amplifier/filter
section 36. Such values will be used to identify lead failure.
[0026] The output of all 19 channels of the amplifier/ filter 36 is fed to the multi-channel
A/D converter 38 which is controlled by an input processor 44 for sampling rate settings.
The analog signals are converted to digital data format suitable for input to the
input processor 44. A/D converters sold by Analog Devices, Norwood MA can be used
for this purpose.
[0027] The multi-channel A/D converter 38 is optically coupled to the input processor 44
by optical isolator 40. All control lines to the A/D convertor 38 are also optically
isolated by optical isolator 42. Any optical isolator can be used for this purpose.
[0028] All DC power lines connected to the amplifiers 36 and A/D converter 38 are also isolated
from the AC power line with a DC/DC convertor 43 in order to provide complete patient
isolation from ground. DC/DC converters available from Burr Brown can be used for
this purpose.
[0029] The basic instructions for controlling operation of the input processor 44 are stored
in a read only memory (ROM) 46. The random access memory (RAM) 48 is used as a buffer
memory for data and a portion of the RAM 48 can also be used as program memory when
a control program is being downloaded from the microcomputer 18. The input processor
44 has a bus 50 to communicate with its RAM 48 and ROM 46 and a separate bus 55 for
communicating with the microcomputer 18.
[0030] The memory architecture of the calculation processor is similar to that of the input
processor. The basic instructions for controlling operation of the calculation processor
52 are stored in a read only memory (ROM) 54. The random access memory (RAM) 56 is
used as a buffer memory for data and a portion of the RAM 56 can also be used as program
memory when a control program is being downloaded from the microcomputer 18. The calculation
processor 52 has a bus 58 to communicate with its RAM 56 and ROM 54 and uses the bus
55 for communicating with the microcomputer 18.
[0031] The A/D converter 38 acquires the data at high speed and filtering is done by the
input processor 44 to exclude frequencies outside the region of interest. The input
processor simultaneously decimates the sampling rate of the input data to a lower
sampling rate. The input processor 44 transfers the filtered and decimated data stream
to the microcomputer 18 for display of the raw input signals via the data bus 55 and
buffers 60 to the microcomputer data bus 40. The input processor 44 also transfers
the data to the calculation processor 52 for calculation of power spectrum and higher-order
spectrum characteristicS of the input signals via a serial communication interface
51. The calculation processor 52 calculates power spectrum and higher-order spectrum
characteristics of the input data and produces diagnostic indices from the calculated
power spectrum and higher-order spectrum data. The input processor can be any general
purpose DSP processor such as the ADSP-2101 sold by Analog Devices, Norwood MA. The
calculation processor is a floating-point DSP processor in the preferred embodiment
such as the TMS320C30 sold by Texas Instruments, Dallas, TX.
[0032] The host or microcomputer 18 of Fig. 1 is shown in greater detail in Fig. 3. The
entire microcomputer system runs under control of a microprocessor 62 with the program
memory stored in ROM 64. The RAM 66 is used for storage of intermediate data. The
storage device 84 can be a Winchester disk or a large block of RAM or any other storage
medium. It is used for storage of clinical information and can be used for archiving
patient data.
[0033] In a preferred embodiment, the microcomputer 18 contains a math coprocessor 70 which
is connected directly to microprocessor 62. The math coprocessor 70 is used for scalar
and graphic calculations. A graphics controller 72 operating under program control
of the microprocessor 62 drives a graphics display 20. An interface port 74 provides
the connection from the microcomputer bus 40 to the user interface device 24. The
user interface device 24 may be a keyboard, a pointing device or a keypad or any combination
of these or similar devices. The interface port 74 can also provide a connection between
the microcomputer and an external evoked potential stimulating device. This connection
will allow the microcomputer to trigger a stimulus or easily identify the onset of
an independently triggered stimulus.
[0034] Operator control of the entire acquisition, analysis and display procedure is controlled
by the user interface device 24 with feedback on the graphics display 20. The data
bus 40 can be used to send control data to the 19 channel data acquisition system
12 (e.g., filtering, gain, sampling rate, start/stop acquisition, perform self diagnostics)
and to receive EEG data from the system, as well as to download program data to the
system. A serial or parallel port 78 is provided to drive a hard copy output device
22 for printing desired diagnostic indices.
[0035] Referring now to Fig. 4, a block diagram of the system operations and the method
of the present invention is described. As mentioned above, the system and method of
the present invention computes dynamic phase and density relations of EEG signals
from a preselected number of leads. Single-valued diagnostic indices are then generated
from the data arrays by using clinically predetermined coefficient arrays. The results
are quantitative indices useful for analyzing cerebral electrical activity as it relates
to, for example, the assessment of depth and adequacy of anesthesia, pain responses
during surgical stress, cerebral ischemia, cerebral hypoxia, level of consciousness,
degree of cerebral intoxication, altered evoked potential responses, and normal or
abnormal cognitive processes that include but are not limited to Alzheimer's disease
and HIV-related dementias.
[0036] The monitor module 402 handles the overall operations of the system via integration
of data and process information from the user interface module 404, acquisition and
raw EEG data management module 406, power spectral, bispectral and higher-order spectral
processing module 408, and the diagnostic index derivation module 410. A detailed
illustration of module 402 can be found in Fig. 5.
[0037] The operator controls and interacts with the system during the course of a procedure
through the user interface and display management module 404. This interaction includes,
but is not limited to, entry of information regarding the patient and type of diagnostic
procedure underway; lead and acquisition settings; continuous display of acquisition
status, lead integrity, and diagnostic indices corresponding to regions probed by
each electrode; and requests for printing and archiving results to the storage device.
Module 404 directly interacts with the monitor module 402. The operations handled
by module 404 can be achieved under a commercially available environment such as Microsoft
Windows.
[0038] The acquisition and raw EEG data management module 406, handles all of the raw EEG
data checking and processing prior to power spectrum, bispectrum, and higher-order
spectrum analysis. This includes, but is not limited to, continuous acquisition of
EEG data and the verification of its integrity; preparation of all unipolar EEG data
for auto/cross power spectral, bispectral, and higher-order spectral processing. Module
406 directly interacts with the monitor module 402. A more detailed description of
module 406 is provided below in connection with Fig. 7.
[0039] The power spectral, bispectral, and higher-order spectral processing module 408 controls
the generation of all data arrays for power distribution, dynamic phase relations,
and power coupling within the EEG. This information can be obtained by computing the
auto/cross power spectrum, bispectrum, and higher-order spectra using either an FFT-based
or parametric-based approach. The tasks performed by this module include, but are
not limited to: Fourier transformation and the generation of power spectra; auto/cross
bispectral density and higher order density generation; auto/cross bicoherence and
higher order coherence generation; auto/cross bispectral real product and higher-order
real product generation; and auto/cross biphase and higher-order phase generation.
Module 408 directly interacts with the monitor module 402. A more detailed description
of module 408 is provided below in connection with Figs. 8 and 9.
[0040] The diagnostic index derivation module 410 generates the data values used in the
diagnostic process. The task includes, but is not limited to, sorting the values in
the frequency band of interest for each of the required power spectrum, bispectrum,
or higher-order spectrum arrays; dividing each of the sorted arrays into bins representing
portions of the distribution histogram of the sorted data (i.e. top 0-5%, top 5-10%
as well as bottom 5%, etc.); summing the values in each bin to create a single number
variable; creating a diagnostic index by multiplying the resultant sorted values from
auto/cross power spectrum, bispectrum, and higher-order spectrum arrays by clinically
predetermined coefficients; and summing all variables that have been multiplied by
a coefficient to create a final univariate diagnostic index. Module 410 directly interacts
with the monitor module 402. A more detailed description of module 410 is provided
below in connection with Fig. 11.
[0041] A schematic of the operation of the monitor module 402 is shown in Fig. 5. In initializing
step 502, the data arrays are filled with the most recent 63 seconds of raw digitized
EEG signals, and the power spectrum, bispectrum, and higher-order spectrum data for
each lead are initialized to zero. The data files required for storage and files containing
data bases required for the computation of diagnostic indices, are also opened in
the initializing step 502.
[0042] In step 504 the system requests the information required to start the acquisition
and diagnostic process from the user via the user interface module 404. This requested
information includes patient descriptive statistics (sex, age, clinical symptoms,
etc.), type of diagnostic procedure to be conducted, the leads used for auto power
spectrum, bispectrum, and higher-order spectrum analysis as well as the leads to be
used for cross power spectrum, bispectrum, and higher-order spectrum analysis.
[0043] In its default mode of operation the system continuously monitors the depth and adequacy
of anesthesia and any pain responses during surgical stress using a default autobispectrum
database. Default band-pass filtering is performed, passing the range 0.5 to 50 Hz;
the default sampling rate is set at 128 samples per second; and the default gain is
set at 5000 for each lead. The following discussion and description of the preferred
embodiments will emphasize autobispectral processing performed on EEGs from specific
electrode sites that best provide depth of anesthesia information. Other modes of
operation will be described more generally.
[0044] According to the international 10/20 electrode system, the 19 EEG signals that can
be acquired using the system are: Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4,
T5, P3, Pz, P4, T6, 01, and 02 (A1 or A2 for reference).
[0045] In order to perform auto power spectrum, bispectrum, and higher-order spectral analysis,
one signal is required. This signal can be measured directly from any of the above
electrodes or it can be synthesized by linearly combining signals from two or more
EEG leads. For example, two analog signals can be subtracted from each other using
a differential amplifier to yield a third signal. The same operation can be performed
on the two digitized signals using numerical subtraction. The auto power spectrum
data will provide information regarding the power distribution within the signal;
the autobispectrum data will provide information regarding deviation from normality,
quadratic nonlinearities and inter-frequency phase relationships within the signal;
finally, auto higher-order spectrum data will provide information regarding deviation
from normality, higher-order nonlinearities, and inter-frequency phase relationships
within the signal. Such processing will determine if the signal is made up of independent
wave components or whether certain frequencies are simply harmonics of nonlinearly
interacting fundamentals. Cerebral phenomena that alter the nonlinear frequency structure
of the signal at the location probed by the electrode are best quantified by autobispectrum
and higher-order spectrum type approaches.
[0046] In order to perform cross power spectrum, bispectrum, and higher-order spectrum analysis,
two signals are required. The two signals can be measured directly from any of the
above electrodes or either of the two signals can be synthesized by linearly combining
two or more of the EEG leads as described earlier. The cross power spectrum data will
provide information regarding the power correlation between the two signals. The cross
bispectrum data will provide information regarding deviation from normality, quadratic
nonlinearities, and inter-frequency phase relationships between the two signals. Finally,
cross higher-order spectrum data will provide information regarding deviation from
normality, higher-order nonlinearities, and inter-frequency phase relationships between
the two signals. Such processing will determine if the frequencies in signal "X" are
independent or whether they are harmonics of fundamentals present in signal "Y". This
provides a better characterization of the relationship between two signals originating
from separate regions of the cortex. Cerebral phenomena that alter nonlinear frequency
relations between the various regions of the cortex are best quantified by cross bispectrum
and cross higher-order spectrum approaches.
[0047] Since the effects of anesthesia are reflected by more global changes in the EEG,
the preferred embodiment will use six signals to illustrate the operation of the system
using autobispectrum analysis for the monitoring of the depth of anesthesia. The Six
signals are derived from the following electrode placements: left and right frontal
(FL/FR) signals are derived from (Fp1-Cz) and (Fp2-Cz) respectively; left and right
parietal (PL/PR) signals are derived from (P3-Cz) and (P4-Cz) respectively; left and
right fronto-parietal (FPL/FPR) signals are derived from (Fp1-P3) and (Fp2-P4) respectively.
[0048] In step 506, a new one-second buffer of raw EEG data is acquired. The system performs
artifact detection on the new one-second buffer and properly updates all data arrays.
Any transmission of artifactual data is displayed to the operator in order to invoke
the operator into correcting the problem.
[0049] The system, in step 508, computes auto power spectrum and autobispectrum arrays for
the signals FL, FR, PL, PR, FPL, FPR. Other signals may, of course, be used for auto/cross
power spectral, bispectral, and higher-order spectral processing. Two different approaches
for power spectrum, bispectrum, and higher-order spectrum computation will be discussed
below with reference to Figs. 8 and 9.
[0050] In step 510, the single-valued diagnostic indices from all generated auto/cross power
spectrum, bispectrum, and higher-order spectrum arrays are computed. The clinically
predetermined coefficient arrays for the auto/cross power spectrum, bispectrum, and
higher-order spectrum arrays are used for the diagnostic index computations. The generation
of the coefficient arrays is discussed later. The system instantaneously displays,
in step 512, all computed diagnostic indices for all signals being analyzed. In step
514, the system checks for an exit request, and if such a request has not been made,
the system repeats steps 508 through 514. In step 516, requested printouts are produced,
results are stored to a storage device for archival purposes and all files are closed.
In step 518, the process is terminated.
[0051] A sample condensed display representation generated by the system is shown in Figs.
6(a) - 6(c). Representations of the patient's head are shown on the graphics display
in Fig 6(a) and Fig. 6(b) . The first illustration Fig. 6(a) is divided into nineteen
sections each representing the region probed by an electrode. The second illustration
Fig. 6(b) is divided into three horizontal sections representing combined left and
right hemisphere activity probed by a group of electrodes in that region. The virtual
head displayed on the screen may be partitioned as required for a particular diagnostic
or monitoring application. For example, if a global effect like depth of anesthesia
is being tracked, then one unified index along with its trend may occupy the whole
display area.
[0052] For head representation Fig. 6(a), each section contains the instantaneous value
of the index 602 using EEG data acquired from the electrode in that region. For head
representation Fig. 6(b), each section contains the instantaneous value of the computed
index 604 using EEG data acquired from several electrodes in that region. Next to
each index value, a color-coded arrow is used to show the instantaneous change in
the direction of the index. The arrow will be green if the index is within acceptable
limits set as by the operator. The arrow will change to yellow if the index moves
into a warning zone. A flashing red bar will replace the arrow if the index has a
value that is outside the acceptable limits set for the patient.
[0053] At the request of the operator, the instantaneous value of the index and its trend
for any section can be displayed as an enlarged view 606 for closer examination as
shown in Fig. 6(c). This will facilitate the examination of the patient's status at
a distance. Each section will be covered by a large "X" 608 if a lead fails or artifact
was detected, for any of the leads contributing to the data required to generate the
diagnostic index for that region.
[0054] Referring to Fig. 7, the acquisition and raw EEG data management module 406 will
now be described in greater detail. In step 702, the system checks whether new data
is being acquired for the first time. If it is, the acquisition system 12 in step
704 is supplied with requested filtering, gain, sampling rate, and lead selection
information. The default settings are band pass 0.5 - 50 Hz for filtering, 5000 for
gain, 128 samples/sec for sampling rate and signals from the lead combinations FL,
FR, PL, PR, FPL and FPR are acquired. The above settings are quite different when
the system is analyzing evoked EEG responses rather than continuous EEG signals. Common
gain and filter settings to acquire signals for the various EEG evoked potentials
are described below.
[0055] EEG evoked potentials are a means by which the sensory areas of the brain and of
the central nervous system may be assayed by detecting responses in the EEG to sensory
stimuli. There are three common methods: Pattern-shift visual evoked potentials (PSVEP)
involve a visual pattern that is shown to the patient and changed. For example, a
strobe light may be flashed or a black and white checkerboard may be reversed (black
for white and vice versa). Brainstem auditory evoked potentials (BAEP) uses a controlled
auditory stimulus such as a click produced by a signal generator. Finally, somatosensory
evoked potentials (SEP) employs either physiologic (touch or muscle stretch) or electrical
stimuli. In all evoked potential methods, electrodes are placed near the appropriate
centers of the brain (i.e. over the visual cortexes in the case of visual evoked potentials)
and EEGs are recorded for a certain period of time beginning with the administration
of the stimulus. The stimulus is repeated many times and the resulting recordings
are averaged (traditionally, in the time domain) so as to eliminate all parts of the
EEG signal except that due to the stimulus. In the present invention, a series of
power spectrum, bispectrum, or higher-order spectrum arrays, as produced from the
EEG of the evoked responses, is averaged.
[0056] For each evoked potential method, different filter and gain settings are used. For
example, a range of common gain settings for pattern-shift visual evoked potentials
is 20,000 to 100,000. A range of common filter settings for PSVEPs is 1 to 3 Hz for
the low end of the band pass and 100 Hz to 300 Hz for the high end. The methods and
use of evoked potentials are described more fully in Evoked Potentials In Clinical
Medicine, by Chiappa 1983.
[0057] In step 706, the acquisition system 12 acquires one second's worth of new data for
all requested leads. Or the signal from one complete evoked potential response is
acquired if the system is analyzing evoked potentials. The system detects lead failures
during the acquisition cycle in step 708 by checking for very large positive or negative
values. Also in step 708, publicly available algorithms are used to check for artifacts
in each lead. In step 710, leads that have failed and those producing artifactual
data are marked for the monitor module 402.
[0058] In step 712, the most recent 4-second record for each of the signals is assigned
to X
i(t), where X
i(t) is the individual time series records provided for auto power spectral, autobispectral,
and auto higher-order spectral processing (herein, the time series X
i(t) (for all t, for one specific i) is referred to as a record). In situations where
cross power spectral, bispectral, and higher-order spectral processing is required,
the most recent 4-second record from the second signal is assigned to Y
i(t). In the preferred embodiment, Y
i(t) is set to equal X
i(t) in all cases, since only auto power spectrum, auto bispectrum, and auto higher-order
spectrum computations are to be performed. The index i denotes the record number from
1 to 60. If evoked potentials are being analyzed, the most recent complete evoked
potential response from each signal is assigned to the appropriate X
i(t) and Y
i(t) as described above. Using evoked potential responses as individual records will
allow us to average a large number of them in the power spectrum, bispectrum, and
higher-order spectrum domains.
[0059] In step 714, a circular buffer mechanism is used for storing the raw EEG for each
lead, as well as the auto/cross power spectrum, bispectrum, and higher-order spectrum
arrays for the sixty most recent 4-second X
i(t) and Y
i(t) records for each lead. The buffer is updated by storing the most recently acquired
and processed data in the location of the oldest data. Operation of the system returns
to the monitor module 402 in step 716.
[0060] Referring now to Fig. 8, the frequency-domain-based procedures for producing the
auto power spectrum, autobispectrum, cross power spectrum, or the cross bispectrum
will now be discussed.
[0061] In step 802, the system checks whether the computation to be performed requires one
signal or two signals. Typically, one time series is required to perform autospectrum
analysis and two time series are required to perform cross spectrum analysis.
[0062] In step 804, the system sets time records in the following manner in order to proceed
with auto power spectral or autobispectral computation of the unipolar lead. As these
computations require only one signal, the second set of records (Y
i(t)) is set to equal the first set (X
i(t)). As a consequence, the corresponding Fourier transforms of X
i(t) and Y
i(t), respectively X
i(f) and Y
i(f), are also equal:
where i denotes the record number which, in this embodiment, ranges from 1 to 60.
[0063] In step 806, time records are set for cross power spectral and cross bispectral analysis
using two separate time series signals. As a consequence, the corresponding Fourier
transforms are not equal:
where X
i(t) and Y
i(t) represent individually derived time series records from two separate regions probed
by two or more electrodes.
[0064] The fast Fourier transform (FFT) X
i(f) and Y
i(f) of each of the 60 selected records for that signal, is computed using a standard
IEEE library routine (or any other publicly available routine) in step 808. If requested,
the series of transformed records, X
i(f) and Y
i(f), may be each normalized by dividing the value at each frequency by the constants
C
xi and C
yi, respectively. These constants are derived separately for each record and each series
(either X or Y). The constant could be the total power, the largest peak in the spectrum
of interest, or some other derivative of X
i(f), X
i(t), Y
i(f), and Y
i(t). In step 810, the System checks whether the computation to be performed is a power
spectrum or bispectrum computation.
[0065] The system computes the auto/cross power spectral density values (PD(f)) in step
812 by using the following equations where PC(f) is the average complex product for
a signal or signal pair:
where
Y(f) is the complex conjugate of Y
i(f) (0 < f < N/2) and M is the number of records (60 in the preferred embodiment).
The system then returns the requested auto/cross power spectral density array to monitor
module 402.
[0066] If the system is performing a bispectral computation in step 814, the system checks
whether the computation to be performed is an autobispectrum or cross bispectrum computation.
[0067] Autobispectrum analysis is a special case of crossbispectrum analysis and therefore
different rules of symmetry apply. In step 816, the system uses the following equations
to determine what ranges of f
1 and f
2 to use during autobispectral computation:
where N equals the number of samples per record (512 = 4 secs per record * 128 samples
per sec in a preferred embodiment), and
where f
1 and f
2 (also referred to as F
1 and F
2 or Frequency 1 and Frequency 2) denote the frequency pairs over which bispectrum
computation will be carried out.
[0068] In step 818, the following equations are used to determine the range of f
1 and f
2 for cross bispectrum analysis:
where all variables represent the same values as they do for autobispectral analysis,
except that for crossbispectral analysis X
i(f) and Y
i(f) represent the Fourier transform of the individually derived time series records
from two separate regions.
[0069] In Step 820, the power spectra P
xi(f) and P
yi(f) of each of the 60 selected records for that signal are computed by squaring the
magnitudes of each element of the Fourier transform X
i(f) and Y
i(f) respectively.
[0070] The system computes the average complex triple product in step 822 by using the following
equations where bc
i(f
1,f
2) is an individual complex triple product from one 4-second record and BC(f
1,f
2) is the average complex triple product for all 60 records:
where
is the complex conjugate of
, and
where M is the number of records (60 in the preferred embodiment)
[0071] The average real triple product is computed in step 824 by using the following equations
where br
i(f
1,f
2) is an individual real triple product from one 4-second record and BR(f
1,f
2) is the average real triple product for all 60 records:
where M is the number of records (60 in the preferred embodiment)
[0072] In step 826, the array of auto/cross bispectral density values (BD(f
1,f
2)) are computed using the following equation:
[0073] In step 828, the array of the square roots of the average real triple products (SBR(f
1,f
2)) are computed using the following equation:
[0074] In step 830, the System computes the array of auto/cross biphase values (ϕ(f
1,f
2)) using the following equation:
[0075] In step 832, the system computes the array of auto/cross bicoherence values (R(f
1,f
2)) using the following equation:
[0076] In step 834, the system returns the requested auto/cross power spectral density array
or auto/cross bispectral density, squared-rooted average real triple product, bicoherence,
biphase arrays to the monitor module 402.
[0077] The above frequency-domain-based equations used to compute the auto/cross bispectrum
arrays can be generalized to compute auto/cross higher-order spectral arrays. This
will allow the computation of the trispectrum, quadspectrum, etc. Assuming that the
arrays for a Kth-order spectrum are to be computed the following equations can be
used:
[0078] The average complex Kth order product:
where M is the number of records (60 in the preferred embodiment)
[0079] The average real Kth-order product:
[0080] The auto/cross Kth-order spectral density:
[0081] The auto/cross Kth-order coherence:
[0082] The auto/cross Kth-order phase:
[0083] Figure 9 illustrates a parametric-based method for producing the auto power spectrum,
autobispectrum, cross power spectrum, or cross bispectrum. In steps 902, 904, and
906 the system sets the time series records in the same manner as described above
in steps 802, 804, and 806 respectively. The auto/cross power spectra of X
i(t) and Y
i(t) are estimated in steps 908, 910, and 912. This estimation method includes two
major stages, the autoregressive (AR) model order selection and the auto/cross power
spectrum computation for X
i(t) and Y
i(t). In step 908, the System computes two sequences of autocorrelations, {R
2X(m)} and {R
2Y(m)} using the following equation.
z = X or Y, and m = 0, 1, ..., L
where M is the number of records of each signal (60 in the described embodiment),
and N is the number of samples per record (512 in the described embodiment), and L
is greater than the largest possible AR filter order (50 in the described embodiment).
[0084] The Final Prediction Errors, FPE
X(m) and FPE
Y(m) are calculated for all orders, m = 1, 2, ..., L, by performing a Levinson recursion
function on each autocorrelation sequence in step 910 in order to find the order of
the AR filter. The order of the AR filters can be determined by finding the location
of the minimum of Final Prediction Errors: FPE
X(m) and FPE
Y(m) respectively, i.e.,
where Q
x and Q
y are the locations of the minimum values for FPE
x(m) and FPE
y(m) (respectively) and, consequently, the orders of the AR filters of the power spectra
X
i(t) and Y
i(t) (respectively).
[0085] Once the orders of the AR filters for auto power spectra are known, the autocorrelation
sequences, {R
2X(m)} and (R
2Y(m)}, are entered into a Levinson recursion with order Q
X and Q
Y, respectively, instead of L. The coefficients, {c
iX, i=0, 1, ...,Q
X} and (c
iY, i = 0,1, ... ,Q
Y), obtained from the recursion are the coefficients of the AR filters for auto power
spectra of X
i(t) and Y
i(t) respectively. Then, in step 912, the transfer function of the AR filters for auto
power spectra of X
i(t) and Y
i(t) are computed as the square root of the prediction error (σ
z) divided by the Fourier transform of the coefficients, i.e.,
[0086] The auto/cross power spectral density values (PD(f)) is the magnitude of the complex
product of
HPX(
f) and the complex conjugate of
HPY(
f), i.e.,
[0087] If requested, the same normalization used in step 808 is may be used here (on
HPz(
f)).
[0088] In step 914, the system checks whether the computation to be performed is a bispectrum
computation, and if it is not, the system returns the requested auto/cross power spectral
density array to monitor module 402.
[0089] In steps 916, 918, and 920, the system sets the symmetries in the same manner as
described above in steps 814, 816, and 818.
[0090] The system estimates the auto/cross bispectrum in steps 922, 924, and 926. The estimation
process includes two major stages: the order selection and bispectrum computation.
In step 922, two sequences of third-order moments, {R
3X(τ)} and {R
3Y(τ)} are computed using the following equation.
where
,
, and L is greater than the largest possible AR filter order (e.g. 50).
[0091] In step 924, two matrices T
X and T
Y are formed as follows.
[0092] From the assumption we made about the AR filter of bispectrum, the orders O
X and O
Y of the AR filters of bispectra of X
i(t) and Y
i(t) are the ranks of the super matrices T
X and T
Y. Therefore, O
X and O
Y are chosen by using singular value decomposition. Having found the orders, we obtain
the coefficients of the AR filters of the bispectra by solving the following linear
system of equations:
where the skewness (β
z) and the coefficients (b
1z, ..., b
O2z),
, can be obtained by solving the linear system of equations.
[0093] The auto/cross bispectrum of X
i(t) and Y
i(t) are computed in step 926 as the cubic root of the triple product of the skewnesses
(β
Xβ
Yβ
Y)
1/3, divided by the triple product of the Fourier transforms of the AR filter coefficients
(H
z(f)), i.e.,
and BR(f
1,f
2) is the real triple product for that same signal:
where the auto power spectra of X
i(t) and Y
i(t), P
X(f) and P
Y(f), are computed by squaring the magnitudes of transfer function of the AR filters
for auto power spectra of X
i(t) and Y
i(t) (
HPX(
f) and
HPY(
f) ) respectively. If requested, the same normalization used in step 808 may be used
here. Similarly, (β
z)
1/3/
Hz(
f) is divided by the square root of the sum of the square of its magnitude for certain
frequency band, its largest peak value, or some similarly derived normalizing constant.
[0094] After obtaining power spectrum and auto/cross bispectrum, the system computes the
bispectral density array, the biphase, the bicoherence, and the square-rooted average
real triple product (RTP) array in step 928 in the same way as in steps 826, 828,
830, and 832. In step 930, the system returns to the monitor module 402 the requested
auto/cross power spectral density array, bispectral density, square-rooted real triple
product, biphase, and bicoherence arrays.
[0095] The above parametric equations used to compute the auto/cross bispectral arrays can
be generalized to compute auto/cross higher-order spectral arrays. This will allow
the computation of the trispectrum, quadspectrum, etc. Assuming that the arrays for
a Kth-order spectrum are to be computed the following equations can be used:
[0096] The auto/cross Kth-order spectrum:
[0097] The real Kth-order product:
[0098] After obtaining the auto/cross Kth-order spectrum, the system computes the auto/cross
Kth-order spectral density array, the auto/cross Kth-order phase, and the auto/cross
Kth-order coherence the same way as in the frequency-domain-based method.
[0099] For illustration purposes Figs. 10(a) - 10(c) are graphs of sample autobispectral
arrays showing frequency pairs 0 < f
1 < 30 Hz, and 0 < f
2 < 15 Hz. A bispectral density array is shown in Fig. 10(a) where the Z axis represents
the magnitude in decibels (db) of the coupled interaction between all appropriate
frequency pairs f
1 and f
2. Recall that the frequency pair (f
1, f
2) must adhere to the equation:
where N = 256 Hz in this case. A bicoherence array is shown in Fig. 10(c) where the
Z axis represents the normalized magnitude in percent (%) of the coupled interaction
between all appropriate frequency pairs f
1 and f
2. A biphase array is shown in Fig. 10(b) where the Z axis represents the phase in
radians of the coupled interaction between all appropriate frequency pairs f
1 and f
2. An array of square root of real triple product is shown in Fig. 10(d) where the
Z axis represents the magnitude in decibels (db) of the coupled interaction between
all appropriate frequency pairs f
1 and f
2.
[0100] Referring to figure 11, a more detailed illustration of the diagnostic index generation
module 410 will now be provided. In step 1102, the system identifies the type of diagnostic
assessment in progress. In a preferred embodiment, the 5 possible options are:
1. Depth of anesthesia, consciousness, pain & surgical stress.
2. Cerebral ischemia and hypoxia.
3. Cerebral intoxication (alcohol, narcotics).
4. Evoked potential evaluation
5. Cognitive process evaluation
[0101] In step 1104, the system gets the auto/cross power spectrum, bispectrum, and/or higher-order
spectrum arrays that are required for the computation of the requested diagnostic
index using the sorting method described below. The various arrays that can be used
in the generation of the diagnostic index are: auto/cross power spectrum; auto/cross
bispectral density; auto/cross bicoherence; auto/cross bispectral real product; auto/cross
biphase; auto/cross Kth-order spectral density; auto/cross Kth-order coherence; auto/cross
Kth-order spectral real product; and auto/cross Kth-order phase;
[0102] The sorting of auto/cross power spectrum, bispectrum, and higher-order spectrum arrays
is an important feature as it provides a mechanism to compensate for changes in the
energy distribution in these (and any other) spectra. The following is a general description
of how the feature is implemented:
[0103] Based on an FFT derived from 4-second records, as described in the preferred embodiment,
120 data points can be computed for a power spectrum array that covers the frequency
band 0-30 Hz (with 4-second records and a sampling rate of 128 samples per second,
the resolution of the FFT is 0.25 Hz, and the range used is 30 Hz wide, thus there
are 120 = 30 Hz / 0.25 Hz data points). When the 120 data points are sorted in descending
order, the first element in the sorted array will correspond to the largest power
spectrum value, and the last element will correspond to the smallest power spectrum
value. A distribution histogram of the power can then be generated using the sorted
array. The X axis on the histogram will represent power in dBs and the Y axis will
represent the number of points in the sorted array that correspond to a particular
X axis power value. If all points in the sorted array are added together, the sum
will represent the total power in the 0-30 Hz spectrum. If a number of adjacent points
in the sorted array are added together, a portion of the histogram representing a
percentage of total power is obtained. For example, in a particular EEG signal, the
top 2 points in the sorted array represents the top 10% of the total power in the
power distribution histogram. Similarly adding the bottom 70 points (for the same
signal) in the sorted array will give the bottom 10% of the total power in the histogram.
Given this approach any portion of the power distribution histogram can be obtained
by adding adjacent elements in the sorted array (top 25% of total power, middle 50%
of total power, etc.) (given that one has empirically determined the transfer function
from specific points to percentage of total power).
[0104] By sorting, we are able to track regions of high activity and low activity (peaks
and valleys) in the 0-30 Hz power spectrum without having to analyze specific narrow
frequency bands. This is equivalent to mapping the power spectrum to its power distribution
function and operating on fixed bands within that distribution function. This transformation
addresses some of the inconsistencies in the behavior of EEG power observed when hypnotic
anesthetic agents are administered. More generally, the sorting scheme outlined above
will transform any auto/cross power spectrum, bispectrum, and higher-order spectrum
array of any dimension, into a one-dimensional distribution function of the values
it contains. The one dimensional distribution is then divided into fixed bands that
can be combined to produce a univariate diagnostic index.
[0105] In step 1106, the reference auto/cross power spectrum, bispectrum, and higher-order
spectrum arrays are sorted. The corresponding dependent arrays are re-ordered according
to the sorted sequence of the reference array. A reference array is an array whose
values are used as the primary sort key for a group of corresponding arrays that have
the same number of variables and are identical in size to the reference array. For
example, if the reference array were to have four elements and they were given the
indices 1, 2, 3, 4 before sorting, and after the sort the new order of the indices
were 2, 1, 4, 3 then one could use the same rearrangement to reorder any other array
of the same size (in this case by placing the second element first, the first element
second, etc.). In this way, one can use the sort of the reference array to rearrange
the dependent arrays. In the preferred embodiment, the reference array is autobispectral
density and the dependent arrays are autobicoherence and the square rooted average
real triple product. Autobispectral density was selected as the reference array because
it provides information about the residual power at each frequency pair after random
phase cancellations. Thus, the sort of the autobispectral density array provides a
more stable means to select autobicoherence and real triple product values than would
the sort of those arrays themselves. A different array may be selected to satisfy
other requirements.
[0106] In step 1108, the sorted auto/cross power spectrum, bispectrum, and higher-order
spectrum arrays are each divided into bins as described earlier. The sum of the points
in each bin for each array is computed and stored in a temporary variable. In step
1110, the clinically predetermined coefficient array for the desired diagnostic index
is retrieved from resident memory (or from the storage device). Each coefficient in
the predetermined coefficient array corresponds to one of the temporary variables
generated in step 1108. In step 1112, the univariate diagnostic index is produced
from the sum of all variables multiplied by their corresponding coefficients in the
predetermined coefficient array. In step 1114, the program returns to the monitor
module 402.
[0107] The predetermined clinical coefficient arrays referred to above are essential to
the device's ability to achieve clinically relevant diagnostic efficacy. The process
adopted for generating these clinical reference arrays will now be described. Since
a large number of possible reference arrays must be generated to accommodate all the
diagnostic modalities of the system, only one will be discussed in detail. All other
reference arrays are generated in a similar fashion. For illustration purposes a method
for generating the coefficients required to track depth of anesthesia using the derived
signals FL and FR (of the preferred embodiment) is described below:
[0108] In order to determine the clinical coefficients for a particular diagnostic index,
raw data as well as clinical diagnoses are required. In the particular case described
below, in order to develop an index which indicates anesthetic depth, EEG signals
and assessments of the patient's response to clinical stimuli were collected. In one
case below, the assessment is based on the patient's change in arterial blood pressure.
In the other case, the assessment is the surgeon's judgment as to whether the patient
had a motor-reflexive response. Once the data have been obtained, the various spectra
and variables as described above may be generated. By performing statistical regressions
on the processed data in conjunction with the clinical diagnoses, the coefficients
which produce the index with the best predictive diagnostic ability can be determined.
[0109] In two separate studies EEG potentials were continuously recorded from a group of
patients undergoing elective surgery. The recording period started at approximately
5 minutes prior to induction and lasted for the duration of the surgery. The derived
signals FL, FR, PL, PR, FPL and FPR were acquired using the procedure described above.
[0110] The purpose of the first study was to determine whether autobispectrum variables
provide information about anesthetic depth at incision. Forty adult patients were
studied. Anesthesia was induced with thiopental (up to 5.0 mg/kg) and intubation performed
after the administration of succinylcholine. Patients were randomly assigned to receive
isoflurane 0.75 MAC (Mean Alveolar Concentration), 1.00 MAC, or 1.25 MAC in 100% oxygen.
End-tidal agent concentration was monitored and after a period of steady-state had
been achieved, purposeful movement in response to skin incision was assessed. Each
patient was classified as either a "mover" or a "non-mover" based on the patient's
response to incision.
[0111] The purpose of the second study was to determine whether autobispectrum variables
provide information about predicting hemodynamic responses to laryngoscopy during
induction with sufentanil or alfentanil. Forty adult patients were studied. Patients
received premedication with oral diazepam (0.05 - 0.15 mg/kg) and were induced with
thiopental (4.0 - 6.0 mg/kg) and 60% nitrous oxide in oxygen, followed by vecuronium
(0.1 mg/kg). Each patient was then randomly assigned to receive one of five dose regimens:
normal saline; alfentanil 15 mcg/kg or 30 mcg/kg; sufentanil 0.5 mcg/kg or 1.5 mcg/kg.
Laryngoscopy was performed 3 minutes after drug administration. Brachial blood pressure
was measured every minute with a cuff device. Patients who exhibited a change in mean
arterial pressure of more than 20% in response to intubation were classified as "responders";
those who did not exhibit such a change at intubation were classified as "non-responders."
[0112] An autobispectral density, an auto bicoherence, and an auto square-rooted average
real triple product array were generated for the derived signals FL and FR for each
patient using a two minute period prior to the stimulus. The frequency band for which
the bispectral arrays were computed was 0.25 - 30 Hz. Each bispectral array contained
3600 data points.
[0113] The resultant auto bispectral density, auto bicoherence, and auto square-rooted average
real triple product arrays were sorted using the auto bispectral density array as
the sorting reference array. The sorting was done using the algorithm described above.
Eleven variables were produced from each of the sorted arrays as described below:
Var1 =Sum of the largest 15 points in sorted array
Var2 =Sum of points ranked 16th to 30th in sorted array
Var3 =Sum of points ranked 31st to 50th in sorted array
Var4 =Sum of points ranked 51th to 100th in sorted array
Var5 =Sum of points ranked 101th to 150th in sorted array
Var6 =Sum of points ranked 151th to 300th in sorted array
Var7 =Sum of points ranked 301th to 500th in sorted array
Var8 =Sum of points ranked 501st to 900th in sorted array
Var9 =Sum of points ranked 901st to 1500th in sorted array
Var10=Sum of points ranked 1501st to 2400th in sorted array
Var11=Sum of points ranked 2401st to 3600th in sorted array
[0114] The values of the 11 variables for each array were computed. As a result, there were
33 temporary variables per patient per signal.
[0115] The 80 patients were then classified into two groups. The first group contained all
the patients from the first study that moved at incision and all the patients from
the second study that had a change in blood pressure of greater than 20% in response
to intubation. The second had all patients from the first study who did not move at
incision and all the patients from the second study that had a blood pressure response
of less than 20% for intubation.
[0116] In order to produce a set of coefficients that would yield the most effective diagnostic
index, a discriminant analysis was performed. The diagnostic index (I(c
0, c
1, . . ., c
33)) for a set of coefficients (c
0, c
1, . . ., c
33) is given by:
where BIS
A through BIS
K are the 11 sorted temporary variables from the bispectrum array; BIC
A through BIC
K are the variables from the bicoherence array; and PS
A through PS
K are the variables from the sorted square-rooted average real triple product array.
The discriminant analysis, given the values of the temporary variables mentioned above
and the responder/non-responder classification for each patient, produces the set
of coefficients which yield the best separation of responders and non-responders by
the function I. Discriminant analysis algorithms are publically available; in this
case, the ones used are from the statistics library available from IMSL (Houston,
Texas). Below is a sample list of coefficients generated using a database of 170 patients:
|
for derived signals FL, FR |
C0 |
-4.28 |
C1 |
-0.65 |
C2 |
+0.57 |
C3 |
+1.21 |
C4 |
-1.23 |
C5 |
+2.63 |
C6 |
-3.34 |
C7 |
+2.11 |
C8 |
+2.74 |
C9 |
-3.08 |
C10 |
0.0 |
C11 |
-0.66 |
C12 |
+0.04 |
C13 |
-1.86 |
C14 |
+0.50 |
C15 |
+0.14 |
C16 |
-0.30 |
C17 |
+0.15 |
C18 |
-0.08 |
C19 |
-0.11 |
C20 |
+0.05 |
C21 |
+0.05 |
C22 |
-0.02 |
C23 |
+0.67 |
C24 |
-1.02 |
C25 |
0.0 |
C26 |
-0.19 |
C27 |
-1.27 |
C28 |
+1.20 |
C29 |
+1.25 |
C30 |
-2.15 |
C31 |
-2.43 |
C32 |
+3.16 |
C33 |
+0.64 |
[0117] For the two studies discussed above the univariate index was used to predict the
response to the stimulus for each patient. The following is a summary of the results
achieved:
- Sensitivity; predicting movement at incision
- = 96%
- Specificity; predicting no movement at incision
- = 63%
- Overall accuracy; predicting move/no move at incision
- = 83%
- Sensitivity; predicting >20% BP change at intubation
- =100%
- Specificity; predicting <20% BP change at intubation
- = 50%
- Overall accuracy; predicting BP change at intubation
- = 85%
[0118] The example above shows one approach to obtaining a set of coefficients for a diagnostic
application in a retrospective manner. Several other approaches can be used to separate
the clinical populations being studied using a univariate index. Such approaches include
but are not limited to linear regression, stepwise linear regression, logistic regression,
and stepwise logistic regression. Of course regardless of which method is used to
retrospectively compute the coefficients, performance of the final index must be confirmed
in a prospective trial prior to using it in patient care.
[0119] The analytic process described above is used to generate the reference databases
for cerebral ischemia, cerebral hypoxia, consciousness, degrees of intoxication, altered
evoked potential responses, and normal or abnormal cognitive processes including but
not limited to identifying patients with Alzheimer's disease and HIV-related dementias.
[0120] In addition to quantifying the depth, adequacy of anesthesia, and pain responses
during surgical stress, the system and method of the present invention may also be
used to assess a myriad of cerebral phenomena that alter the nonlinear frequency structure
of the EEG as quantified by bispectrum and higher-order spectrum approaches. Such
cerebral phenomena include but are not limited to, cerebral ischemia, cerebral hypoxia,
level of consciousness, degree of cerebral intoxication, altered evoked potential
responses, and normal or abnormal cognitive processes caused by neurological disorders
like Alzheimer's disease or HIV-related dementias.
[0121] Although power spectrum and bispectrum analysis techniques have been applied to the
EEG signal for diagnostic purposes, as was discussed in the background above, higher-order
spectral approaches have never been used. Furthermore no power spectrum, bispectrum,
or higher-order spectrum technique has ever been used in conjunction with the sorting
method described above. Specifically, the system and method sort various auto/cross
power spectrum, bispectrum, and higher-order spectrum arrays, divides the sorted arrays
into bins, sums the variables in each bin, multiplies the value from each bin by a
clinically derived coefficient, and finally adds all variables together to generate
a univariate diagnostic index. The different arrays that can be used are: auto/cross
power spectrum, auto/cross bispectral density, auto/cross bicoherence, auto/cross
biphase, auto/cross average real triple product, auto/cross Kth-order spectral density,
auto/cross Kth-order coherence, auto/cross phase, and auto/cross real product.
[0122] While the foregoing invention has been described with reference to its preferred
embodiments, various alterations and modifications will occur to those skilled in
the art. All such alterations and modifications are intended to fall within the scope
of the appended claims.
1. A method of noninvasively detecting cerebral phenomena comprising the steps of:
acquiring electroencephalographic signals through at least one electrode (13) from
a body surface of a subject (15) being analysed;
band pass filtering said electroencephalographic signals to obtain filtered signals
in a desired frequency range and
dividing said filtered signals into a plurality of data records
characterised in that the method comprises the steps of characterising dynamic phase
relations within said filtered signals by processing said filtered signals to generate
Kth-order spectral values, where K is an integer greater than 2 and
deriving from said generated Kth-order spectral values a diagnostic index that quantifies
the detected cerebral phenomena.
2. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the step
of acquiring electroencephalographic signals further comprises the step of attaching
electrodes (13) to a head (14) of the subject (15) being analysed in order to obtain
a unipolar electroencephalographic signal from each region of interest of both left
and right hemispheres of the subject's brain to which said electrodes are attached.
3. The method of noninvasively detecting cerebral phenomena of claim 1 wherein said Kth-order
spectral values generated in said step of characterising said dynamic phase relations
are auto Kth-order spectral density values.
4. The method of noninvasively detecting cerebral phenomena of claim 3 wherein said auto
Kth-order spectral density values are auto Kth-order phase values.
5. The method of noninvasively detecting cerebral phenomena of claim 3 wherein said auto
Kth-order spectral density values are auto Kth-order coherence values.
6. The method of noninvasively detecting cerebral phenomena of claim 1 wherein said step
of acquiring electroencephalographic signals further comprises the step of attaching
electrodes (13) to the head (14) of the subject (15) being analysed in order to obtain
bipolar data sets of electroencephalographic signals from left and right hemispheres
of the subject's brain to which said electrodes are attached.
7. The method of noninvasively detecting cerebral phenomena of claim 6 wherein one bipolar
data set is acquired from a frontal left hemisphere of the subject's brain and another
bipolar data set is acquired from a frontal right hemisphere of the subject's brain.
8. The method of noninvasively detecting cerebral phenomena of claim 6 wherein one bipolar
data set is acquired from a left occipital region of the subject's brain and another
bipolar data set is acquired from a right occipital region of the subject's brain.
9. The method of noninvasively detecting cerebral phenomena of claim 6 wherein one bipolar
data set is acquired from a left parietal region of the subject's brain and another
bipolar data set is acquired from a right parietal region of the subject's brain.
10. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomena being detected are pain responses during surgical stress in the subject
(15) being analysed.
11. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomenon being detected is the degree of cerebral intoxication of the subject (15)
being analysed.
12. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomena being detected are normal or abnormal cognitive processes of the subject
(15) being analysed.
13. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomenon being detected is chronic ischemia or infarction in the subject (15) being
analysed.
14. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomenon being detected are cognitive processes caused by neurological disorders.
15. The method of noninvasively detecting cerebral phenomena of claim 1 wherein the cerebral
phenomenon being detected are altered evoked potential responses.
16. A system for noninvasively detecting cerebral phenomena comprising:
means (406) for acquiring electroencephalographic signals through at least one electrode
(13) from a body surface of a subject (15) being analysed; means (406) for band pass
filtering said electroencephalographic signals to eliminate those signals outside
the desired frequency range and
means (406) for dividing said filtered signals into a plurality of data records;
characterised in that the system comprises means (408) for generating Kth-order spectral
values capable of characterising dynamic phase relations within said filtered electroencephalographic
signals where K is an integer greater than 2 and
means (410) for deriving a diagnostic index from said generated Kth-order opectral
values that quantifies the detected cerebral phenomena.
17. The system for noninvasively detecting cerebral phenomena of claim 16 further comprising
a plurality of said means (406) for acquiring electroencephalographic signals being
connected to said means (406) for filtering.
18. The system for noninvasively detecting cerebral phenomena of claim 17 wherein said
plurality of said means (406) for acquiring electroencephalographic signals is a plurality
of electrodes (13) attachable to a head (14) of a subject (15) being analysed to obtain
a unipolar electroencephalographic signal from each of a plurality of regions of interest
on both left and right hemispheres of the subject's brain.
19. The system for noninvasively detecting cerebral phenomena of claim 16 wherein said
means (406) for acquiring electroencephalographic signals comprises:
a pluralitiy of surface electrodes (13) for mounting on a surface of a head (14) of
the subject (15) being analysed; means (32) for providing defibrillator protection
for limiting voltage to said amplifier during a discharge;
means (36) for amplifying said filtered signals for a high gain in order to maximize
the dynamic range for high frequency, low energy wave components of said filtered
signals;
means (30) for providing electrosurgery protection for limiting radio frequency current
through said means for amplifying;
means for feeding said signals to an analog-to-digital converter (38) to convert said
signals to digital signals.
20. The system for noninvasively detecting cerebral phenomena of claim 16 wherein means
(406) for acquiring electroencephalographic signals is adapted to aquire signals from
different regions of a brain of said subject.
21. The method of noninvasively detecting cerebral phenomena of claim 1, wherein said
step of deriving from said generated Kth-order spectral values a diagnostic index
comprises the steps of:
sorting said spectral values into predetermined bins of ranges of spectral values;
summing all of the spectral values in each bin;
multiplying the sum of spectral values in each bin by a predetermined coefficient
to obtain a bin product;
summing said bin products to obtain a diagnostic index which represents a degree of
presence or absence of said phenomena.
22. The system for noninvasively detecting cerebral phenomena of claim 16, wherein the
means for deriving a diagnostic index from said Kth-order spectral values comprises:
means for sorting said spectral values into predetermined bins of ranges of spectral
values;
means for summing all of the spectral values in each bin;
means for multiplying the sum of spectral values in each bin by a predetermined coefficient
to obtain a bin product;
means for summing said bin products to obtain a diagnostic index which represents
a degree of presence or absence of said phenomena.
23. The method of noninvasively detecting cerebral phenomena of claim 1 wherein said step
of acquiring electroencephalographic signals comprises acquiring electro-encephalographic
signals through at least one electrode placed near areas of the brain in which an
evoked potential response occurs.
24. The method of noninvasively detecting cerebral phenomena of claim 1 wherein said step
of acquiring electroencephalographic signals comprises acquiring electro-encephalographic
signals from one complete evoked potential response.
25. The method of noninvasively detecting cerebral phenomena of claim 1, wherein said
Kth-order spectral values are third order spectral values.
26. The system of noninvasively detecting cerebral phenomena of claim 16, wherein said
Kth-order spectral values are third order spectral values.
1. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene, welches die folgenden
Schritte umfaßt:
Erfassen elektroencephalographischer Signale über mindestens eine Elektrode (13) von
einer Körperoberfläche eines zu analysierenden Objektes (15);
Bandpaßfiltern der genannten elektroencephalographischen Signale, um gefilterte Signale
in einem gewünschten Frequenzbereich zu erhalten und
Aufteilen der genannten gefilterten Signale in eine Mehrzahl von Datensätzen
dadurch gekennzeichnet, daß das Verfahren die Schritte des Charakterisierens dynamischer
Phasenrelationen innerhalb der genannten gefilterten Signale durch Verarbeiten der
genannten gefilterten Signale zum Erzeugen spektraler Werte K-ter Ordnung, worin K
eine ganze Zahl größer als 2 ist, umfaßt und
Ableiten eines diagnostischen Index, der die detektierten zerebralen Phänomene quantifiziert
von den genannten erzeugten spektralen Werten K-ter Ordnung.
2. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
der Schritt des Erfassens elektroencephalographischer Signale weiters den Schritt
des Anbringens von Elektroden (13) auf dem Kopf (14) eines zu untersuchenden Subjektes
(15) umfaßt, um ein unipolares elektroencephalographisches Signal von jedem interessierenden
Bereich von sowohl der linken als auch rechten Hemisphäre des Hirnes des Subjektes
zu erhalten, an das die Elektroden angebracht sind.
3. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
die spektralen Werte K-ter Ordnung, die im genannten Schritt des Charakterisierens
der genannten dynamischen Phasenrelationen erzeugt werden, autospektrale Dichtewerte
K-ter Ordnung sind.
4. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 3, worin
die genannten autospektralen Dichtewerte K-ter Ordnung Autophasenwerte K-ter Ordnung
sind.
5. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 3, worin
die genannten autospektralen Dichtewerte K-ter Ordnung Autokohärenzwerte K-ter Ordnung
sind.
6. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
der genannte Schritt des Erfassens elektroencephalographischer Signale weiters den
Schritt des Anbringens von Elektroden (13) am Kopf (14) des zu untersuchenden Subjektes
(15) umfaßt, um bipolare Datensätze von elektroencephalographischen Signalen von der
linken und rechten Hemisphäre des Hirnes des Subjektes zu erhalten, an das die Elektroden
angebracht sind.
7. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 6, worin
ein bipolarer Datensatz von einer frontalen linken Hemisphäre des Hirnes des Subjektes
erfaßt wird und ein anderer bipolarer Datensatz von einer frontalen rechten Hemisphäre
des Hirnes des Subjektes erfaßt wird.
8. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 6, worin
ein bipolarer Datensatz von einer linken occiputalen Region des Hirnes des Subjektes
erfaßt wird und ein anderer bipolarer Datensatz von einer rechten occiputalen Region
des Hirnes des Subjektes erfaßt wird.
9. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 6, worin
ein bipolarer Datensatz von einer linken parietalen Region des Hirnes des Subjektes
und ein anderer bipolarer Datensatz von einer rechten parietalen Region des Hirnes
des Subjektes erfaßt wird.
10. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
die zerebralen detektierten Phänomene Schmerzantworten während chirurgischem Streß
am zu analysierenden Subjekt sind.
11. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
das detektierte zerebrale Phänomen der Grad der zerebralen Intoxikation des zu analysierenden
Subjektes (15) ist.
12. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
die detektierten zerebralen Phänomene normale oder abnormale kognitive Prozesse des
zu analysierenden Subjektes (15) sind.
13. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
das detektierte zerebrale Phänomen chronische Ischämie oder ein Infarkt im zu analysierenden
Subjekt (15) ist.
14. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
das detektierte zerebrale Phänomen in kognitiven Prozessen besteht, die durch neurologische
Störungen hervorgerufen sind.
15. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
das detektierte zerebrale Phänomen in veränderlichen hervorgerufenen Potentialantworten
besteht.
16. System zum nichtinvasiven Detektieren zerebraler Phänomene umfassend:
Mittel (406) zum Erfassen elektroencephalographischer Signale durch mindestens eine
Elektrode (13) von einer Körperoberfläche eines zu analysierenden Subjektes (15);
Mittel (16) zum Bandpaßfiltern der genannten elektroencephalographischen Signale,
um jene Signale außerhalb eines gewünschten Frequenzbereiches zu eliminieren und
Mittel (406) zum Aufteilen der genannten gefilterten Signale in eine Vielzahl von
Datensätzen,
dadurch gekennzeichnet, daß das System Mittel (408) zum Erzeugen spektraler Werte
K-ter Ordnung umfaßt, die Fähig sind, dynamische Phasenrelationen in den genannten
gefilterten elektroencephalographischen Signalen zu charakterisieren, worin K eine
ganze Zahl größer als 2 ist, und
Mittel (410) zum Ableiten eines diagnostischen Index aus den genannten spektralen
Werten K-ter Ordnung, um die zerebralen Werte zu quantifizieren.
17. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 16, welches
weiters eine Mehrzahl von genannten Mitteln (406) zum Erfassen elektroencephalographischer
Signale umfaßt, die an die genannten Mittel (406) zum Filtern angeschlossen sind.
18. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 17, worin
die genannten Vielzahl der genannten Mittel (406) zum Erfassen elektroencephalographischer
Signale eine Vielzahl von Elektroden (13) ist, die an den Kopf (14) eines zu analysierenden
Subjektes (15) anschließbar sind, um ein unipolares elektroencephalographisches Signal
von jeder der Vielzahl der interessierenden Regionen auf der linken und rechten Hemisphäre
des Hirnes des Subjektes zu erhalten.
19. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 16, worin
die genannten Mittel (406) zum Erfassen elektroencephalographischer Signale umfassen:
eine Mehrzahl von Oberflächenelektroden (13) zum Befestigen an einer Oberfläche eines
Kopfes (14) eines zu untersuchenden Subjektes (15);
Mittel (32) zum Bereitstellen eines Defibrilatorschutzes zum Begrenzen der Spannung
zum Verstärker während einer Entladung;
Mittel (36) zum Verstärken der gefilterten Signale für eine hohe Verstärkung, um den
dynamischen Bereich für hochfrequente niederenergetische Wellenkomponenten der genannten
gefilterten Signale zu maximieren;
Mittel (30) zum Bereitstellen eines elektrochirurgischen Schutzes zum Begrenzen des
Radiofrequenzstromes durch die genannten Verstärkermittel;
Mittel zum Liefern der genannten Signale zu einem Analog-Digital-Wandler (38), um
die genannten Signale in digitale Signale zu wandeln.
20. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 16, worin
die Mittel (406) zum Erfassen elektroencephalographischer Signale ausgebildet sind,
um Signale von verschiedenen Regionen des Gehirns des genannten Subjektes zu erfassen.
21. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
der genannte Schritt des Ableitens eines diagnostischen Index aus den genannten spektralen
Werten K-ter Ordnung die Schritte umfaßt:
Sortieren der genannten spektralen Werte in vorbestimmte Ablagen von Bereichen der
spektralen Werte;
Aufsummieren aller spektralen Werte in jeder Ablage;
Multiplizieren der Summe des spektralen Wertes in jeder Ablage mit einem vorbestimmten
Koeffizienten, um ein Ablage-Produkt zu erhalten;
Aufsummieren der genannten Ablage-Produkte um einen diagnostischen Index zu erhalten,
der den Grad des Vorhandenseins oder Nichtvorhandenseins der genannten Phänomene repräsentiert.
22. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 16, worin
die Mittel zum Ableiten eines diagnostischen Index aus den genannten spektralen Werten
K-ter Ordnung umfassen:
Mittel zum Auffeilen der genannten spektralen Werte in vorbestimmte Ablagen von Bereichen
der spektralen Werte;
Mittel zum Summieren aller spektralen Werte in jeder Ablage;
Mittel zum Multiplizieren der Summe der spektralen Werte in jeder Ablage mit einem
vorbestimmten Koeffizienten, um ein Ablage-Produkt zu erhalten;
Mittel zum Summieren der genannten Ablageprodukte, um einen diagnostischen Index zu
erhalten, der den Grad des Vorhandenseins bzw. Nichtvorhandenseins der genannten Phänomene
repräsentiert.
23. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
der genannten Schritt des Erfassens elektroencephalographischer Signale das Erfassen
elektroencephalographischer Signale über mindestens eine Elektrode umfaßt, die in
der Nähe jener Bereiche des Gehirns angeordnet sind, in denen eine hervorgerufene
Potentialantwort auftritt.
24. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
der genannte Schritt des Erfassens elektroencephalographischer Signale das Erfassen
elektroencephalographischer Signale von einer vollständigen hervorgerufenen Potentialantwort
umfaßt.
25. Verfahren zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 1, worin
die spektralen Werte K-ter Ordnung spektrale Werte dritter Ordnung sind.
26. System zum nichtinvasiven Detektieren zerebraler Phänomene nach Anspruch 16, worin
die genannten spektralen Werte K-ter Ordnung spektrale Werte dritter Ordnung sind.
1. Procédé de détection non invasive des phénomènes cérébraux comprenant les étapes :
- d'acquisition de signaux électroencéphalographiques par l'intermédiaire d'au moins
une électrode (13) à partir d'une surface corporelle d'un sujet (15) soumis à l'analyse
;
- de filtrage passe-bande desdits signaux électroencéphalographiques afin d'obtenir
des signaux filtrés dans une gamme de fréquences souhaitée ;et
- de division desdits signaux filtrés en une pluralité d'enregistrements de données
;
caractérisé en ce que le procédé comprend les étapes de caractérisation des relations
de phase dynamiques dans lesdits signaux filtrés en traitant lesdits signaux filtrés
pour générer des valeurs spectrales d'ordre K, où K est un entier supérieur à 2 ;
et de détermination à partir desdites valeurs spectrales générées d'ordre K, d'un
indice de diagnostic qui quantifie les phénomènes cérébraux détectés.
2. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel l'étape d'acquisition de signaux électroencéphalographiques comprend en
outre l'étape de fixation d'électrodes (13) à la tête (14) d'un sujet (15) soumis
à l'analyse afin d'obtenir un signal électroencéphalographique unipolaire de chaque
région intéressante, des deux hémisphères gauche et droit du cerveau du sujet auquel
sont fixées lesdites électrodes.
3. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel lesdites valeurs spectrales d'ordre K produites lors de ladite étape de
caractérisation desdites relations de phase dynamiques sont des valeurs d'auto-densité
spectrale d'ordre K.
4. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 3,
dans lequel lesdites valeurs d'auto-densité spectrale d'ordre K sont des valeurs d'auto-phase
d'ordre K.
5. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 3,
dans lequel lesdites valeurs d'auto-densité spectrale d'ordre K sont des valeurs d'auto-cohérence
d'ordre K.
6. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel ladite étape d'acquisition de signaux électroencéphalographiques comprend
en outre l'étape de fixation d'électrodes (13) à la tête (14) du sujet (15) soumis
à l'analyse afin d'obtenir des jeux de données bipolaires de signaux électroencéphalographiques
provenant des hémisphères gauche et- droit du cerveau du sujet auquel sont fixées
lesdites électrodes.
7. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 6,
dans lequel un jeu de données bipolaires est acquis à partir d'un hémisphère frontal
gauche du cerveau du sujet et un autre jeu de données bipolaires est acquis à partir
de l'hémisphère frontal droit du cerveau du sujet.
8. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 6,
dans lequel un jeu de données bipolaires est acquis à partir d'une région occipitale
gauche du cerveau du sujet et un autre jeu de données bipolaires est acquis à partir
d'une région occipitale droite du cerveau du sujet.
9. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 6,
dans lequel un jeu de données bipolaires est acquis à partir d'une région pariétale
gauche du cerveau du sujet et un autre jeu de données bipolaires est acquis à partir
d'une région pariétale droite du cerveau du sujet.
10. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel les phénomènes cérébraux détectés sont des réponses à la douleur pendant
un stress chirurgical chez le sujet (15) soumis à l'analyse.
11. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel le phénomène cérébral détecté est le degré d'intoxication cérébrale du
sujet (15) soumis à l'analyse.
12. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel les phénomènes cérébraux détectés sont des processus cognitifs normaux
ou anormaux du sujet (15) soumis à l'analyse.
13. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel le phénomène cérébral détecté est une ischémie ou un infarctus chronique
chez le sujet (15) soumis à l'analyse.
14. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel les phénomènes cérébraux détectés sont des processus cognitifs provoqués
par des désordres neurologiques.
15. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel les phénomènes cérébraux détectés sont des réponses altérées à un potentiel
évoqué.
16. Système pour détecter de façon non invasive des phénomènes cérébraux, comprenant :
- des moyens (406) pour acquérir des signaux électroencéphalographiques par l'intermédiaire
d'au moins une électrode (13) à partir d'une surface corporelle d'un sujet (15) soumis
à l'analyse ;
- des moyens (406) pour soumettre à un filtrage passe-bande lesdits signaux électroencéphalographiques
pour éliminer les signaux se situant à l'extérieur de la gamme de fréquences souhaitée
; et
- des moyens (406) pour diviser lesdits signaux filtrés en une pluralité d'enregistrements
de données ;
caractérisé en ce que le système comprend des moyens (408) pour générer des valeurs
spectrales d'ordre K capables de caractériser des relations de phase dynamiques au
sein desdits signaux électroencéphalographiques filtrés, où K est un entier supérieur
à 2 ; et
- des moyens (410) pour déterminer un indice de diagnostic à partir desdites valeurs
spectrales générées d'ordre K, qui quantifie les phénomènes cérébraux détectés.
17. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
16, comprenant en outre une pluralité desdits moyens (406) destinés à acquérir des
signaux électroencéphalographiques qui sont connectés auxdits moyens (406) de filtrage.
18. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
17, dans lequel ladite pluralité desdits moyens (406) destinés à acquérir des signaux
électroencéphalographiques sont constitués d'une pluralité d'électrodes (13) pouvant
être fixées à la tête (14) d'un sujet (15) soumis à l'analyse pour obtenir des signaux
électroencéphalographiques unipolaires à partir de chacune d'une pluralité de régions
étudiées sur les deux hémisphères gauche et droit du cerveau du sujet.
19. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
16, dans lequel lesdits moyens (406) destinés à acquérir des signaux électroencéphalographiques
comprennent :
- une pluralité d'électrodes de surface (13) destinées à être montées sur une surface
de la tête (14) du sujet (15) soumis à l'analyse ;
- des moyens (32) pour assurer une protection vis-à-vis d'un défibrillateur afin de
limiter la tension appliquée audit amplificateur pendant une décharge ;
- des moyens (36) pour amplifier lesdits signaux filtrés avec un gain élevé afin de
rendre maximale la gamme dynamique pour les composantes d'ondes à haute fréquence
et à basse énergie desdits signaux filtrés ;
- des moyens (30) pour assurer une protection vis-à-vis de l'électrochirurqie afin
de limiter le courant à fréquence radio passant dans lesdits moyens d'amplification
;
- des moyens pour appliquer lesdits signaux à un convertisseur analogique-numérique
(38) afin de convertir lesdits signaux en signaux numériques.
20. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
16, dans lequel des moyens (406) destinés à acquérir des signaux électroencéphalographiques
sont adaptés à acquérir des signaux provenant de différentes régions du cerveau dudit
sujet.
21. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
1, dans lequel ladite étape de détermination à partir desdites valeurs spectrales
d'ordre K générées d'un indice de diagnostic, comprend les étapes :
- de tri desdites valeurs spectrales dans des classes prédéterminées de gammes de
valeurs spectrales ;
- de sommation de toutes les valeurs spectrales dans chaque classe ;
- de multiplication de la somme de valeurs spectrales dans chaque classe par un coefficient
prédéterminé afin d'obtenir un produit de classe ;
- de sommation desdits produits de classes afin d'obtenir un indice de diagnostic
qui représente un degré de présence ou d'absence desdits phénomènes.
22. Système pour détecter de façon non invasive des phénomènes cérébraux selon la revendication
16, dans lequel les moyens destinés à déterminer un indice de diagnostic à partir
desdites valeurs spectrales d'ordre K comprennent :
- des moyens pour trier lesdites valeurs spectrales dans des classes prédéterminées
de gammes de valeurs spectrales ;
- des moyens pour sommer toutes les valeurs spectrales dans chaque classe ;
- des moyens pour multiplier la somme de valeurs spectrales dans chaque classe par
un coefficient prédéterminé afin d'obtenir un produit de classe ;
- des moyens pour sommer lesdits produits de classes afin d'obtenir un indice de diagnostic
qui représente un degré de présence ou d'absence desdits phénomènes.
23. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel ladite étape d'acquisition de signaux électroencéphalographiques comprend
l'acquisition de signaux électroencéphalographiques au moyen d'au moins une électrode
placée à proximité de zones du cerveau dans lesquelles il se produit une réponse à
un potentiel évoqué.
24. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel ladite étape d'acquisition de signaux électroencéphalographiques comprend
l'acquisition de signaux électroencéphalographiques à partir d'une réponse complète
à un potentiel évoqué.
25. Procédé de détection non invasive de phénomènes cérébraux selon la revendication 1,
dans lequel lesdites valeurs spectrales d'ordre K sont des valeurs spectrales d'ordre
trois.
26. Système de détection non invasive de phénomènes cérébraux selon la revendication 16,
dans lequel lesdites valeurs spectrales d'ordre K sont des valeurs spectrales d'ordre
trois.